Is pancreatic cancer the same as pancreatic ductal adenocarcinoma? Almost always: about 80 percent of pancreatic cancers are ductal adenocarcinomas (Cancer Research UK types page), and this record describes that disease. The rarer types (acinar cell carcinoma, the cystic tumours and the carcinomas that arise in them, solid pseudopapillary neoplasm, pancreatoblastoma) and the neuroendocrine tumours of the islets each have a page linked from this one and behave differently.
How common is it? 531,318 new cases and 490,786 deaths a year worldwide (GLOBOCAN 2024), 11,479 cases and about 10,200 deaths a year in the UK, where it is the 10th commonest cancer but the 5th commonest cause of cancer death (Cancer Research UK), and 67,530 cases and 52,740 deaths projected in the United States for 2026, where it is the third leading cause of cancer death (SEER). The UK lifetime risk is about 1 in 55 to 1 in 59.
Is it getting more common? Yes. UK incidence rates are up 21 percent since the early 1990s and 7 percent in the last decade, and the count is projected to reach about 16,000 a year by 2038 to 2040 (Cancer Research UK); incidence rose in 14 of 48 countries for men and 17 for women over a decade (Huang 2021); and in the United States it is projected to be the second leading cause of cancer death by 2040 (Rahib 2021). Ageing populations, obesity and diabetes are the usual explanations; mortality rates have not fallen because treatment gains have been small until recently.
What raises the risk? Age above all (median 71 in the United States). Smoking doubles the risk (2.2 times for current smokers, back to normal 20 years after quitting) and causes 22 percent of UK cases; obesity carries about 1.5 times the risk and causes 12 percent; type 2 diabetes about 1.8 to 1.9 times; chronic pancreatitis about 8 times at five years; a first-degree relative with the disease 1.6 to 1.8 times; alcohol at 3 or more units a day 15 to 19 percent more (Bosetti 2012; Genkinger 2011; Huxley 2005; Ben 2011; Kirkegard 2017; Cancer Research UK). About 5.5 percent of patients carry an inherited variant in ATM, BRCA2, BRCA1, CDKN2A, TP53 or MLH1, most without a family history (Hu 2018; Shindo 2017).
I have just been told I have diabetes. Should I worry? For most people, no: about 1 in 100 adults who develop diabetes after 50 is diagnosed with pancreatic cancer within three years (Chari 2005). But new diabetes with weight loss rather than weight gain, or diabetes that suddenly becomes hard to control, is a recognised warning sign: NICE NG12 lists new-onset diabetes with weight loss at 60 or over as a reason for an urgent CT, and research scores such as ENDPAC use weight change, glucose change and age to pick out the small group who need a scan (Sharma 2018).
Why is it found so late? The early tumour causes no symptoms; when symptoms come they are vague (back pain, indigestion, weight loss) unless the tumour blocks the bile duct and causes jaundice, which is why head tumours are found earlier than body and tail tumours. In England in 2019, 45 percent of cases were diagnosed after an emergency presentation and only 22 percent through an urgent suspected cancer referral; only about a quarter are stage I or II at diagnosis (Cancer Research UK). In the SYMPTOM study no first symptom distinguished people who turned out to have cancer from those who did not (Walter 2016).
Which tests will I have? A pancreatic protocol CT first (before any stent if you are jaundiced), then PET-CT and or endoscopic ultrasound with a needle sample if the diagnosis is unclear, blood tests including CA 19-9, and, if surgery is possible, sometimes MRI of the liver or a keyhole look inside the abdomen (NICE NG85 1.1 and 1.3). A biopsy is often not needed before surgery when the scan is clear (Cancer Research UK). Everyone should be offered a genetic (germline) test.
What do resectable, borderline resectable and locally advanced mean? They describe how the tumour sits against the arteries and veins behind the pancreas on the CT scan, and they decide whether surgery comes first, after chemotherapy, or not at all: resectable means no arterial contact and limited venous contact; borderline resectable means arterial contact of less than 180 degrees or venous involvement the surgeon can reconstruct (or a very high CA 19-9, or uncertain spread, or poor fitness); locally advanced means the tumour encases an artery or blocks a vein beyond repair (Isaji 2018; NCCN). The stage number (1 to 4) is a separate description of size, nodes and spread.
What does an R1 margin on my pathology report mean? In the UK, cancer cells within 1 mm of the cut edge of the removed tissue, even if not touching it (Royal College of Pathologists; Campbell 2009). Reported this way, most head-of-pancreas resections are R1 (76 to 85 percent in the studies that introduced the rule) because the tumour spreads along nerves and vessels behind the gland; it does not mean the operation failed, and adjuvantchemotherapy is given whatever the margin. Direct involvement of a margin carries a worse outlook than involvement within 1 mm (Ghaneh 2019).
Should my relatives be screened? There is no NHS screening programme for the general population and none is recommended (UK National Screening Committee; the US task force advises against it). Surveillance with yearly MRI/MRCP or endoscopic ultrasound is offered to people with hereditary pancreatitis and a PRSS1 mutation, to carriers of BRCA1, BRCA2, PALB2 or CDKN2A who have an affected first-degree relative, and to Peutz-Jeghers syndrome, and considered for two or more affected first-degree relatives across two generations or Lynch syndrome with an affected relative (NICE NG85 1.1.15 to 1.1.17). In the CAPS programmes 77.8 percent of cancers found under surveillance were stage I and five-year survival was 73.3 percent (Dbouk 2022). Ask the team for a referral to genetics; the trial record for PRECEDE and the EUROPAC registry are linked from this page.
Why have I been given enzyme capsules? Most pancreatic cancers block the duct that carries digestive enzymes, so fat is not absorbed and weight falls. NICE offers enteric-coated pancreatin to everyone with unresectable disease and considers it before and after surgery (NG85 1.6); in UK primary care records only 21.7 percent of patients received it, and those who did lived longer in a matched comparison (Roberts 2019). Fish oils are not recommended for weight loss (NG85 1.6.3).
What is the outlook? Averages hide a wide range. Across everyone in the United States 13.7 percent are alive at five years; for the 15 percent found while confined to the pancreas it is 43.6 percent, and resected patients who complete modern adjuvantchemotherapy do better still (the treatment rows carry those trial figures); for distant disease it is 3.4 percent (SEER). In the UK about 25 percent survive one year and about 5 percent ten years (Cancer Research UK). These figures predate the 2026 RASinhibitor results and are population averages, not a personal prognosis.
The science in detail, 1 of 7 (the molecular layer; the headline is on the overview). Genomic landscape. KRAS mutation in 88 to 94% (G12D about 40%, G12V about 32%, G12R about 16%, Q61 about 7%, G12C 1 to 2%), TP53 66 to 76%, CDKN2A loss 37 to 48% with CDKN2B and often MTAP, SMAD4 loss 17 to 33%, RNF43 5 to 8%, ARID1A 5 to 9%, KDM6A 3 to 4%, GNAS R201 2 to 4%, MYC amplification 4 to 13%. The 6 to 12% of KRAS wild-type tumours carry the fusions and alternative drivers instead: BRAF mutation 13% and fusion 6.6% of them, FGFR2 5.2%, ALK 2.6%, NRG1 1.3%, RET 1.3%, with MSI-high 4.7% against 0.7%. Germline: BRCA2 1.4 to 2%, ATM 1.2 to 2.3%, BRCA1 0.4 to 1%, PALB2 0.2 to 0.6%, any susceptibility gene 4 to 10% and no family history in most carriers. Homologous recombination deficiency by gene mutation 11 to 19%; mismatch repair deficiency 1 to 2%; TMB 10 or more, 1 to 2%. Every figure and cohort is in the molecular table; the cBioPortal rows were computed on TCGA, QCMG, ICGC, UTSW, CPTAC and the two 2024 MSK cohorts.
The science in detail, 2 of 7 (the molecular layer; the headline is on the overview). Transcriptional subtypes. Two survive every re-analysis: classical (GATA6-high) and basal-like or squamous (GATA6-low). In COMPASS, basal-like was 20% of 195 advanced patients, with response to first-linechemotherapy 10% against 33%, progression on modified FOLFIRINOX 60% against 15% and median overall survival 5.9 against 9.3 months; a GATA6 in situ hybridisation stain calls the subtype with sensitivity 89% and specificity 83%. Collisson's three, Bailey's four (squamous, progenitor, immunogenic, ADEX) and Puleo's five map onto the same axis; the exocrine-like and ADEX classes were acinar contamination. Subtype is a continuum set by mutant KRAS and GATA6 copy number, about 12% of tumours are hybrids with intermediate survival, and squamous histology is the tissue form of the basal-like signature, often as a subclone.
The science in detail, 3 of 7 (the molecular layer; the headline is on the overview). Precursors. More than 99% of the smallest PanIN-1lesions already carry KRAS, CDKN2A, GNAS or BRAFmutations; grading is now two-tier (Baltimore consensus 2015). IPMNs carry GNAS R201 in 66% and RNF43 in most, mucinous cystic neoplasms RNF43 without GNAS, and both are direct precursors, with SMAD4 and TGFBR2 loss marking the invasive step and about three years between high-grade dysplasia and cancer. A cancer beside a cyst is related in 51% of cases and independent in 18%. The classical KRAS-CDKN2A-TP53-SMAD4 ladder is only part of the story: two-thirds of genomes show punctuated, mitotic-error rearrangements that can knock out several drivers at once, and precursor cells travel along the ducts.
The science in detail, 4 of 7 (the molecular layer; the headline is on the overview). Stroma and immunity. The desmoplastic stroma limits perfusion, but removing it is harmful: depleting myofibroblasts or hedgehog signalling gave more aggressive tumours in mice, and hyaluronidase raised the response rate without changing survival in phase 3. Fibroblasts exist as myofibroblastic, inflammatory and antigen-presenting states, and tissue organises into reactive (immune-hot) and deserted (chemoprotective) neighbourhoods. T cells are excluded by CXCL12 from FAP-positive fibroblasts, and MHC class I is degraded by autophagy. Every checkpoint trial outside mismatch repair deficiency has failed: ipilimumab 0 of 27, anti-PD-L1 0 of 14, durvalumab with or without tremelimumab 3.1% and 0% of 65. The exceptions point the way: long-term survivors have both many high-quality neoantigens and abundant CD8 T cells, and those clones are edited out at metastasis.
The science in detail, 5 of 7 (the molecular layer; the headline is on the overview). Monitoring. CA 19-9has pooled sensitivity 79% and specificity 82%, rises with cholestasis, and cannot be made at all by the roughly 10% of patients who are Lewis-negative (FUT3-null), whose medianlevel is 2.4 against 496 U/mL and whose survival matches the worst marker group; a value of 7 U/mL or below identifies them with 95% positive predictive value. Baseline level is prognostic but a 50% fall on treatment is not a valid surrogate for survival. ctDNA is found in 43 to 62% of localised disease before surgery and 37 to 49% after it, predicts recurrence with 90% sensitivity about three months before imaging, and grades risk by allele fraction; most assays read only KRAS codons 12, 13 and 61, and no approval uses it.
The science in detail, 6 of 7 (the molecular layer; the headline is on the overview). Early detection, research only. CAPS surveillance of high-risk individuals shifted 57.9% of detected cancers to stage I with median overall survival 9.8 against 1.5 years; the consensus starts at 50 or ten years before the youngest affected relative with annual endoscopic ultrasound and MRI. PRECEDE is building the international cohort and biomarker validation under a PROBE design. Blood tests are not ready: CA 19-9 turns up about two years before diagnosis, multi-analyte panels reach 64% sensitivity in resectable disease at 99.5% specificity, and the methylation test reads 16.8% sensitivity at stage I. New-onset diabetes is the strongest clinical enrichment, with about 1% of people over 50 diagnosed within three years and the ENDPAC score giving a 4.4-fold enrichment.
The science in detail, 7 of 7 (the molecular layer; the headline is on the overview). Testing, in practice. Germlinepanel testing at diagnosis for every patient (BRCA1, BRCA2, PALB2, ATM, CDKN2A, TP53 and the mismatch repairgenes), tumour sequencing with RNA for fusions where KRAS reads wild-type, mismatch repair immunohistochemistry especially for medullary and colloid tumours, allele-level KRAS reporting because the alleles differ in prognosis and in which inhibitor applies, and GATA6 or a subtype classifier where a trial requires it. Homologous recombination scores, ctDNA and TROP2-style expression readouts have no pancreatic label.
Frequencies marked cBioPortal were computed from the public API on 24 September 2026 on the sequenced sample lists of paad_tcga_pan_can_atlas_2018 (184 samples), paad_qcmg_uq_2016 (383 sequenced), paad_icgc (99), paad_utsw_2015 (109), paad_cptac_2021 (140), pdac_msk_2024 (2,336) and pancreas_msk_2024 (395), as sample-level counts of non-synonymous mutation, high-level amplification, deep deletion or structural variant; they are not the papers' own percentages, which are quoted alongside.
KRASallele shares are counts of mutation records over all KRAS mutation records in a study, so a tumour with two KRAS mutations counts twice; the MSK study's own KRAS_VARIANT attribute gives the same ordering. The TCGA deposit reads KRAS in 65% because it keeps the low-cellularity and non-ductal samples the 2017 paper excluded; purified or microdissected cohorts read 92 to 96%.
Panel and exome figures are not interchangeable: copy-number amplification calls run lower on the MSK panel than on TCGA or UTSW arrays and exomes, and the panel TMB median (about 3 per megabase) is three times the exome median.
Machine-readable versions
The same record for scripts and assistants; checked 2026-09-24. Data CC BY-NC 4.0, attribute “Data from OnCo (onco.cc)”.